Where in the AI stack enterprise outcomes get determined is not where capital has been flowing. An op-ed argues that roughly $2 trillion in software value has been vaporized over the past year because investors treated model capability as the core competitive question, while Salesforce has signaled that the actual battleground sits at a different layer.
The cost of the wrong question
The constraint is one of framing. When markets price AI competition as a foundation-model race, software company valuations follow model-adjacency rather than the economics of enterprise deployment. That is the thesis error the op-ed identifies. The $2 trillion in erased software value is, by this reading, not a correction on AI fundamentals but on which question the market was asking.
The piece argues that model performance is not the variable that drives software sector outcomes. Salesforce, the op-ed contends, has made visible where competition actually forms. It is not the model layer. The implication is that a year of capital allocated against that frame is the cost of misidentifying the battleground before Salesforce pointed elsewhere.